Was only testing for handle less than next_handle, but not checking for
negative handle values. Fixed to use buffer_handle_valid() to test for a
valid handle.
- added keymap_setup() function to keymap tests
- The test string was not long enough to overflow the larger buffer.
Switched to use pseudo-random sequence intstead of a fixe test string.
factor out common code and add back wrappers, so keyboard initializes to
default keyboard on power up, and afterwards, won't reinitialize the
keyboard unless the map changes.
- update ARCHITECTURE-keymaps.org file for new keymap definition method
- remove unused keymap table template files
- in keymap_setup.c template, set default to keymap 0.
- Added the apple 2 CAPS map
- auto-generate a function to check validity of a keymap index
- Added buffered message printing. This is different from keycode
buffering, since an extra delay is added for each message character to
allow polling hosts to keep up. Keycodes are generated at human speeds
and need no further slowdown.
- Added a message character delay to the arch-* modules
- enlarged the buffer pool, and created a buffer for messages
- bumped version number
Accessing library functions via function pointer seems to cause a reset.
Still investigating the cause.
- compile keymaps subdir as part of main src cmakefile, rather than as a
separate library.
- rather than setup table of function pointers and using an indirection
in the main code, set up a jump table (switch). This is cleaner.
This appears to have been a cut-paste typo in the asdf_arch_out2_set() function
in asdf_arch_atmega328p.c. The port was set to OUT2, but the bit was OUT1.
C Flags set in src and test directories were not getting set because
they were assigned to C_FLAGS, which is ignored. Changed to CFLAGS,
which is used by cmake to set default c compilation options.
The encoder boards (ATmega328P, Atmega2560, Apple2(2560) can accomodate
keyboard matrices with no diodes, by adding per-row diodes on the
encoder board. For keyboards with per-key diodes (the typical use
case), the per-row diodes are not needed, and therefore the footprints
should be shunted by default, using a special shunted diode footprint.
When diodes are installed, the shunt can be cut with an x-acto knife.
Previous board revs had the correct footprint, but somehow the regular
diode footprint was substituted on the last release of each of these
PCBS. The correct footprint has been restored.
Dave
Lars Brinkhoff